Diseño del divisor de potencia

The fact that congenital anomalies could be increased in ART pregnancies is plausible for several reasons including:

1. Selective mechanisms operating in vivo against morphologically abnormal sperm might not be comparably operative in vitro.

2. Altered hormonal milieu in vitro may predispose to damage to meiosis or mitosis and hence chromosomal aneuploidy.

3. Point mutations could result from the various chemical exposures inherent during extracorporeal fertilisation.

Although surveys and registries have been set up to record these possible anomalies a recurrent problem is that comparisons are being made to outcomes recorded in birth defects registries, not necessarily the best comparison group. Although so far results have been reassuring (see commentary in Chapter 1) the problem is that disparate rates between centres suggests a lack o f consistent criteria and data recording.

5.3.1 Spuriously increased rates

Examination following ART is often more rigorous than that following conventional pregnancy (controls). Unfortunately this well meaning medical surveillance can be scientifically misleading. Some anomalies such as small umbilical hernias,

pigmented skin lesions and accessory auricles would not necessarily have been recorded in a birth defects registry. It therefore follows that the ART anomaly rates could be spuriously overestimated. A better comparison such as a prospective control study might involve analysis that takes into account the differing standards of record keeping.

If aggressive ART surveillance is not taken into account the reported anomaly rate might be regarded un-critically as increased (spuriously) compared to the less vigilantly assessed general population. Estimates of anomalies in the latter are usually derived from birth defects registries, which record all major anomalies amongst neonates bom in selected hospitals. Although there are several internationally respected registries o f this type other less regimented data collection can lead to underassessment of actual rates. Examples o f the more vigorous surveillance for ART pregnancies include the UK MRC surveillance (Beral and Doyle, 1990). This recorded many anomalies not usually detected in neonatal surveys such as mild hypospadias and undescended testes.

Another assessment bias for ART offspring reflects the more detailed ultrasound scan, usually performed in these pregnancies. For example in the 1990 AFS/SA RT cohort (The American Fertility Society, 1992) probable abnormalities such as periventricular cyst, hydronephrosis, unilateral kidney agenesis, pyloric stenosis and pulmonary hypoplasia were noted. Such anomalies may not present until well after the first week of life, if ever. Finally a subtler source o f bias is that continuing assessments are usually performed on the children after the early neonatal period. These can further contribute to the ascertained incompatibility with normal birth

registries, by continuing to add anomalies to a list, effectively closed in a standard registry.

5.3.2 Spuriously Decreased Rates

5.3.2.1 Failure to look systematically for congenital abnormalities

Bias may also lead to lower than expected ART anomaly rates. In particular, an increased anomaly rate would not be recognised if ART centres did not systematically assess offspring for congenital anomalies. As an example, review of hospital case notes would be a poor way of ascertaining anomalies. Casual surveillance could explain the low anomaly rate recorded in certain surveys, for example 0.7% in Greek cases recorded by Cohen et al, or 0.9 per cent in the 1990 US AFS/ SA ART cohort. The failure to search systematically for anomalies is a valid concern in almost all ART cohorts, with the notable exception of the small IVF case-control study of Morin at al (1989) and the ICSI cohort in Brussels (Bonduelle et al, 1998).

5.3.2.2 Completeness of follow up

Another pitfall arises whenever a significant proportion of a given sample is lost to follow-up. One cannot assume that cases lost to follow-up are comparable to cases remaining for analysis. Moreover, direction of bias is uncertain. Women lost to follow up might not wish to be troubled because their offspring were healthy and because no personal benefit seemed to accrue from participating in a research protocol. The net effect would be a spuriously increased anomaly rate following ART because of the exclusion of normal outcomes. Conversely, women lost to follow up who have experienced a bad outcome (pregnancy loss, or birth of

anomalous infant), may resent the invasion o f privacy or in some way blame an ART centre and therefore fail to respond to requests for follow-up. If so the anomaly rate would be spuriously decreased.

If only a small proportion o f cases from a large cohort are lost to follow up, the remaining subjects could be assumed to be representative. However, this assumption is not necessarily valid when the outcome sought is infrequent or when a relatively high proportion of cases is lost to follow up. In ART surveillance the number of cases lost to follow up may be several fold greater than the background 4 to 5% anomaly rate. In the 1993 US AFS/SA ART report (Society for Assisted Reproductive Technology, 1993) 58 anomalies were reported in 3,873 babies (1.5%), yet 447 o f the babies were lost to follow up (11.4%). If even 10% of the 447 lost cases were anomalous, overall anomaly rates would have risen to 2.7%. In the most recent HFEA report (1998) it was claimed that only 0.8% o f all children bom from IVF were bom with an abnormality. However this figure was obtained by relying on centres recording and reporting this information. Often children are bom away from their fertility treatment centre thus making the likelihood o f such records being valid even more remote.